论文部分内容阅读
应用机器学习方法处理机器阅读的相关任务是人工智能的长远目标,但通常需要大量的人工监督操作.研究一种无监督学习在机器阅读的一个主要任务-语义分析中的应用,这种无监督方法得益于统计关系学习统一框架-Markov逻辑网.鉴于该方法通过依存句法信息无法解析语义分析中普遍存在的反义词、词形变化等语言现象,该文融合WordNet进行改进,促进概念的抽取及合并,并将机器阅读的主要目标-问答作为评价手段,结果表明这种WordNet词典与无监督机器学习相结合的方法可更好地进行语义分析,并且问答正确率可提高至90.6%.
It is a long-term goal of artificial intelligence to deal with the tasks related to machine reading by using machine learning methods, but usually requires a lot of manual supervision.Studying the use of unsupervised learning in semantic analysis, one of the main tasks of machine reading, The method benefits from the unified framework of statistical relations learning - Markov logic network. Since this method can not resolve linguistic phenomena such as antonyms and inflectional forms that are common in semantic analysis, this method integrates WordNet to improve and facilitate the extraction of concepts and And the main objective of machine reading - QA is used as the evaluation method. The result shows that this method of combining WordNet with unsupervised machine learning can be better semantic analysis, and the correct rate of question and answer can be increased to 90.6%.